Generation of three-dimensional digital model of replacement tooth

Abstract

A dental method includes receiving a selection of a replacement tooth of a patient. The method further includes receiving one or more three-dimensional digital tooth models describing one or more teeth of the patient. The method further includes receiving a generated three-dimensional digital model of the replacement tooth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of the patient into the tooth model generation neural network. The tooth model generation neural network includes an encoder portion and a decoder portion. The encoder portion is configured to output a collective feature vector describing one or more digital tooth models. The tooth model generation neural network further includes at least one fully connected layer configured to output a potential spatial vector to the decoder portion in response to receiving the collective feature vector.

Description

Technical Field

[0001] The invention relates to a dental technology, in particular to a modeling of replacement teeth. Background Art

[0002] Dental restorations, such as crowns, are typically prepared to match the patient's existing teeth. To ensure this, a skilled dental technician will sculpt the crown to match the existing teeth and ensure that it fits in the spaces between its adjacent teeth and matches its opposing teeth. Summary of the invention

[0003] The invention provides a dental method, a computer program and a dental system in the independent claims. Implementation examples are given in the dependent claims.

[0004] Correctly modeling replacement teeth for a subject can be challenging. The shape of the replacement teeth should match the subject's existing teeth. A specific implementation example provides an improved method for processing a generated three-dimensional digital model of replacement teeth. It can be implemented by using a tooth model generation neural network that receives one or more three-dimensional digital tooth models as input. The one or more three-dimensional digital tooth models received as input are descriptions of one or more teeth of the subject. For example, they may be a result obtained by scanning the subject's teeth. Using a three-dimensional digital tooth model describing one or more teeth of the subject as input can have the following advantages: the three-dimensional digital model of the replacement teeth generated by the tooth model generation neural network matches the subject's existing teeth. The tooth model generation neural network includes an encoder part and a decoder part. The encoder part generates a collective feature vector, which is converted into a latent space vector of the decoder part through one or more fully connected layers. One advantage of this is that the features of one or more three-dimensional digital tooth models are combined by using the collective feature vector, which is then converted into a latent space vector.

[0005] One aspect of the present invention is to provide a dental method. The method includes receiving a selection of replacement teeth from a subject. The selection of replacement teeth may include providing an identifier or label for the replacement teeth. For example, this may be achieved by providing a name for the replacement tooth or, for example, by providing a number for the identified tooth according to a standard tooth number. For example, a scan of a subject's dental arch, such as an intraoral scan, may include teeth to be replaced, such as partially missing or damaged teeth or prepared teeth, such as dental posts that are ready to receive a dental restoration, such as a crown. The selection of replacement teeth may include identifying the replacement teeth by identifying the teeth to be replaced. For example, the selection of replacement teeth may include indirectly identifying the replacement teeth, such as by identifying adjacent teeth and / or opposing teeth and / or teeth on the same jaw opposite side of the replacement teeth when scanning one or both dental arches of the subject. The replacement tooth may, for example, be a missing tooth. In other examples, the replacement tooth may be a tooth that has been ground down in preparation for receiving, for example, a crown or other dental restoration.

[0006] The method also includes receiving one or more three-dimensional digital tooth models describing one or more teeth of the subject. The one or more three-dimensional digital tooth models are single models. Such a single model may only simulate a specific tooth of the subject. For example, each of the one or more three-dimensional digital tooth models only simulates a single tooth. In this step, the received one or more three-dimensional digital tooth models are used to provide data about the specific teeth of the subject. These one or more three-dimensional digital tooth models can be obtained, for example, by scanning the teeth of the subject (e.g., intraoral scanning), thereby providing data about the geometry and / or aesthetics of the teeth of the subject. These data can help generate a three-dimensional digital replacement tooth model that is personalized to the subject, which matches the geometry and / or aesthetics of the existing teeth in the subject's mouth, such as the subject's existing native teeth and / or artificial teeth.

[0007] The method also includes receiving a generated three-dimensional digital model of the replacement teeth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of the subject into the tooth model generation neural network. The tooth model generation neural network includes an encoder portion and a decoder portion. The encoder portion is configured to output a collective feature vector describing one or more digital tooth models. The tooth model generation neural network also includes at least one fully connected layer, which is configured to output a latent space vector in response to receiving the collective feature vector. The decoder portion is configured to output the generated three-dimensional digital model of the replacement teeth in response to receiving the latent space vector. In this implementation example, the one or more three-dimensional digital tooth models are input into the tooth model generation neural network, and the generated three-dimensional digital model of the replacement teeth is output. This may be beneficial because it can provide an accurate method to generate a three-dimensional digital model of replacement teeth that matches the subject's existing teeth well.

[0008] One or more teeth of the subject may include one or more natural teeth. One or more teeth of the subject may all be natural teeth. That is, they may be the native teeth of the subject. Natural teeth are organic teeth that are naturally developed by the body and are not artificial or processed. The organic parts of natural teeth include crown enamel and dentin, root cementum and dentin, and encapsulated pulp. One or more teeth of the subject may include one or more dentures. One or more teeth of the subject may all be artificial teeth. Artificial teeth are teeth made or manufactured for placement in the mouth of the subject. Such artificial teeth are particularly useful for replacing natural teeth of the subject that are missing or damaged and need to be replaced. A collective feature vector is a vector that describes the features of one or more three-dimensional digital tooth models. In contrast, a latent space vector is a vector that controls how a tooth model generation neural network generates a three-dimensional digital model. The fully connected layer in the neural network connects all inputs to each activation unit of the next layer, that is, each neuron in one layer is connected to each neuron in another layer. The collective feature vector and the latent space vector are converted using at least one fully connected layer. This use of at least one fully connected layer as a conversion between the collective feature vector and the latent space vector may be beneficial because it may enable the individual parts of the tooth model generation neural network to be trained independently. For example, this may generate a higher quality three-dimensional digital model and reduce the amount of data required to generate a good result such as a high quality three-dimensional digital model.

[0009] In one implementation example, the dental method may be used to generate a subject-specific, ie, subject-individualized, three-dimensional digital model of at least partially missing teeth of a subject.

[0010] The tooth model generating neural network may be trained to construct training data by receiving a separate three-dimensional digital model, e.g., a three-dimensional model in the form of a tooth mesh having an original scan of a complete set of teeth of a subject. The complete set of teeth may be, for example, a complete set of teeth of a mandibular dental arch, a complete set of teeth of a maxillary dental arch, or a complete set of teeth of a dentition, i.e., a mandibular and a maxillary dental arch. Training data may be constructed from these complete sets of teeth, e.g., in the form of a tooth mesh, by selecting, for each of these complete sets of teeth, one tooth from the corresponding complete set of teeth as a replacement tooth and using the selected tooth as a reference ground truth. During training, other teeth from the corresponding complete set of teeth may be used as input to the tooth model generating neural network.

[0011] In another implementation example, the dental method further includes selecting the tooth model generation neural network for generating the replacement tooth from a tooth model generation neural network set trained for a specific tooth or a specific tooth type using the selection of the replacement tooth. In this implementation example, the tooth model generation neural network is specially trained to replace a specific tooth or a specific type of tooth. This may be beneficial because the tooth model generation neural network is more focused and can produce better results with less training. Therefore, the tooth model generation neural network set may include generation neural networks for generating different tooth models of different teeth or different types of teeth. The tooth model generation neural network for generating the replacement tooth may be selected from the set based on the specific replacement tooth to be generated and / or based on the specific replacement tooth type to be generated. In some implementation examples, selecting the tooth model generation neural network may determine selecting one or more three-dimensional digital tooth models as inputs to the selected tooth model generation neural network. For example, when training a specific tooth model generation neural network, a specific tooth may be used as input for training. In some cases, the method may include selecting one or more three-dimensional digital tooth models in response to the selection of the tooth model generation neural network.

[0012] When a tooth model generating neural network is trained for a specific type of teeth, it may be, for example, a model specifically for reproducing molars, premolars, canines, incisors, or other types of teeth. Thus, the set of tooth model generating neural networks may, for example, include: a tooth model generating neural network trained for reproducing molars, a tooth model generating neural network trained for reproducing premolars, a tooth model generating neural network trained for reproducing canines, and / or a tooth model generating neural network trained for reproducing incisors.

[0013] When training a tooth model generation neural network for a specific tooth, it can be trained for a specific one of the 32 teeth that are usually included in the dentition of an adult. The 32 teeth include 16 maxillary teeth and 16 mandibular teeth. The 16 maxillary teeth include two contralateral maxillary central incisors, two contralateral maxillary lateral incisors, two contralateral maxillary canines, two contralateral maxillary first premolars, two contralateral maxillary second premolars, two contralateral maxillary first molars, two contralateral maxillary second molars, and two contralateral maxillary third molars. The 16 mandibular teeth include two contralaterally arranged mandibular central incisors, two contralaterally arranged mandibular lateral incisors, two contralaterally arranged mandibular canines, two contralaterally arranged mandibular first premolars, two contralaterally arranged mandibular second premolars, two contralaterally arranged mandibular first molars, two contralaterally arranged mandibular second molars and two contralaterally arranged mandibular third molars. The third molars are commonly known as wisdom teeth and usually erupt between the ages of 17 and 25. The third molars may also never erupt in the oral cavity or may not form at all. When they are formed, they are usually removed. Therefore, the set of tooth model generation neural networks may, for example, include 32 tooth model generation neural networks, each of which is trained to reproduce one of the 32 tooth models typically contained in adult teeth. If there are no third molars, the set of tooth model generation neural networks may therefore include, for example, 28 tooth model generation neural networks. Therefore, the set of tooth model generation neural networks may, for example, include 16 tooth model generation neural networks, each of which is trained to reproduce one of the 16 tooth models typically contained in an adult maxillary dental arch. If there is no third molar, the set of tooth model generation neural networks may include, for example, 14 tooth model generation neural networks. Therefore, the set of tooth model generation neural networks may, for example, include 16 tooth model generation neural networks, each of which is trained to reproduce one of the 16 tooth models typically contained in an adult mandibular dental arch. If there is no third molar, the set of tooth model generation neural networks may include, for example, 14 tooth model generation neural networks.

[0014] In another implementation example, the encoder portion includes a tooth-specific context encoder neural network for each of the one or more three-dimensional digital tooth models. The tooth-specific context encoder neural network is configured to output a tooth feature vector in response to receiving a three-dimensional digital tooth model of the one or more three-dimensional digital tooth models. The tooth model generation neural network is further configured to form a collective feature vector by connecting the tooth feature vectors of at least several of the tooth-specific context encoder neural networks.

[0015] In this implementation example, a specific context encoder neural network is included, which is used for each of the one or more three-dimensional digital tooth models. That is, for each tooth, a three-dimensional digital tooth model is received, the three-dimensional digital tooth model is input into its specific context encoder neural network, and then output as a tooth feature vector. The tooth feature vector describes the specific model of one or more three-dimensional digital tooth models. Then the tooth feature vectors are connected, and then the connection is converted into the latent space vector through at least one fully connected layer. When a specific number of three-dimensional digital tooth models are used as input for generating a three-dimensional digital model of replacement teeth, such as two models, three models or more models, the encoder part may include a tooth-specific context encoder neural network for each of these three-dimensional digital tooth models, such as two tooth-specific context encoder neural networks, three tooth-specific context encoder neural networks or more tooth-specific context encoder neural networks. For example, the encoder part includes a tooth-specific context encoder neural network for each tooth or each type of tooth in the adult dentition. For example, the encoder part may include more tooth-specific context encoder neural networks than the number of input three-dimensional digital tooth models (i.e., one or more three-dimensional digital tooth models). In this case, the encoder part may, for example, include a tooth-specific context encoder neural network for each tooth or each type of tooth included in the adult dentition, and one or more three-dimensional digital tooth models, which are input into their own specific context encoder neural networks based on the teeth or tooth types described by the respective three-dimensional digital tooth models.

[0016] This implementation example may be particularly beneficial because it is an efficient and accurate method for producing a generated three-dimensional digital model of the replacement teeth so that the replacement teeth match the subject, for example, matching the geometry and / or aesthetic features of the subject's existing teeth.

[0017] In another implementation example, the context encoder neural network is a surface-based encoder. The surface-based encoder may have a beneficial effect, that is, a three-dimensional mesh can be directly processed using a surface-based method.

[0018] In another implementation example, the context encoder neural network is a diffusion model encoder (DiffusionNet encoder).

[0019] In another implementation example, the context encoder neural network is a graph-based encoder. The graph-based encoder may have a beneficial effect that a three-dimensional mesh can be directly processed using a graph-based method.

[0020] In another implementation example, the context encoder neural network is a graph neural network.

[0021] In another implementation example, the context encoder neural network is a point-based encoder. The point-based encoder may have a beneficial effect, that is, point clouds can be directly processed using point-based methods. The point cloud is, for example, the raw output of a 3D scanner.

[0022] In another implementation example, the context encoder neural network is a PointNet encoder.

[0023] In another embodiment, the context encoder neural network is a voxel-based encoder.

[0024] In another implementation example, the context encoder is a three-dimensional shape encoder (a three-dimensionalShapeNet encoder).

[0025] In another implementation example, the context encoder neural network is a three-dimensional convolutional neural network.

[0026] In another implementation example, the context encoder neural network is a method based on multi-view rendering.

[0027] In another implementation example, the context encoder neural network is a two-dimensional convolutional neural network.

[0028] In another implementation example, the context encoder neural network is an implicit surface-based encoder. The implicit surface-based encoder may have a beneficial effect, that is, the implicit surface-based method represents a three-dimensional shape of infinite resolution through the implicit surface without requiring a large amount of memory.

[0029] In another implementation example, the context encoder neural network is an occupancy network.

[0030] In another implementation example, the context encoder neural network is the encoder portion of a convolutional autoencoder.

[0031] In another embodiment, the one or more teeth of the subject include one or more adjacent teeth of the replacement tooth. This embodiment is beneficial because by acquiring data of one or more adjacent teeth, the neural network can generate an accurate replacement tooth. The structure and appearance of the adjacent teeth may generally be similar to the replacement tooth. The one or more teeth of the subject described by the one or more three-dimensional digital tooth models may, for example, include an adjacent tooth of the replacement tooth. For example, the replacement tooth may be the last tooth or the first tooth of the dental arch, for example, a second molar in the absence of an adjacent third molar. For example, the adjacent tooth may be selected only if the adjacent tooth is of the same type as the replacement tooth, and the replacement tooth may be a first molar, a first or second premolar, or a lateral incisor. For example, two or more three-dimensional digital tooth models describing two or more teeth of the subject may be received. For example, the teeth of the subject described by the two or more three-dimensional digital tooth models may include two adjacent teeth of the replacement tooth. For example, adjacent teeth may be selected regardless of their tooth type.

[0032] In another embodiment, the one or more teeth of the subject include a tooth on the same jaw as the replacement tooth. This embodiment may be particularly beneficial because the structure and appearance of the tooth on the same jaw as the replacement tooth should be very similar. This will make the replacement or duplication of the replacement tooth more accurate.

[0033] In another embodiment, the one or more teeth of the subject include a set of teeth or a partial dental arch. In this embodiment, a model of teeth that make up a set or a partial dental arch is used. This can provide more data for more accurate modeling of the replacement teeth.

[0034] In another embodiment, one or more teeth include one or more opposing teeth of the replacement tooth. Taking into account symmetry, based on the symmetry of the upper and lower dental arches, each tooth has a symmetrical tooth in the opposite dental arch, i.e., the opposing dental arch, which is also called the primary opposing tooth. For example, one or more teeth may include the primary opposing tooth of the replacement tooth. In addition, the primary opposing teeth of the adjacent teeth of the replacement tooth need to be considered. These opposing teeth can also be called secondary opposing teeth of the replacement tooth. For example, in addition to the primary opposing teeth of the replacement tooth, one or more secondary opposing teeth of the replacement tooth can also be included. For example, in addition to the primary opposing teeth of the replacement tooth, the teeth on the same jaw opposite side of the primary opposing teeth of the replacement tooth can also be included. In addition, one or more teeth on the same jaw opposite side of the secondary opposing teeth of the replacement tooth can also be included. Taking into account the contact between the teeth, two opposing teeth in contact with one tooth can be identified. From the contact perspective, the primary opposing tooth is the tooth with the largest contact area with the corresponding tooth. In the maxilla, the secondary antagonist is located mesial to the primary antagonist, while in the mandible, the secondary antagonist is located distal to the primary antagonist. For example, in terms of contact between the teeth, one or more teeth may include a primary antagonist replacing a tooth as well as a secondary antagonist. Using one or both of them as input to a neural network allows accurate modeling of the replacement tooth.

[0035] In another embodiment, one or more three-dimensional digital tooth models include tooth coordinates describing a position within the subject's mouth, for example, within one of the subject's dental arches and / or within one of the subject's jaws. The one or more three-dimensional digital tooth models are individual models of individual teeth. The selection of replacement teeth may also be marked or may include information about the position of the particular tooth within the subject's mouth. The one or more three-dimensional digital tooth models may be obtained, for example, by scanning the subject's mouth. The scan may obtain descriptive data of the hard and soft tissue structures within the subject's mouth, such as teeth and gums. The data may in particular describe a position within the subject's mouth and may be used to provide tooth coordinates of each tooth.

[0036] For example, one or more teeth of the subject include one or more adjacent teeth of the replacement tooth and / or one or more teeth of the subject include one or more opposing teeth of the replacement tooth. Specifying the adjacent teeth and / or opposing teeth can provide information about the position of the replacement tooth. The tooth generation neural network is, for example, also configured to output the position coordinates of the replacement tooth in response to receiving one or more three-dimensional digital tooth models as input. This implementation example may be beneficial because a method for automatically positioning the generated three-dimensional digital model in a subject's oral model can be provided by providing the position coordinates.

[0037] In another implementation example, the tooth model generation neural network is configured to output the position coordinates of the replacement teeth. In this implementation example, the tooth model generation neural network is directly configured to output the position coordinates of the replacement teeth. For example, the neural network may encode the position coordinates. This may be beneficial because it provides the position coordinates of the generated three-dimensional digital model at the same time.

[0038] In another implementation example, the decoder part is configured to output the position coordinates. In this implementation example, the same decoder part used to generate the generated three-dimensional digital model of the replacement tooth also outputs the position coordinates. This may be a convenient way to automatically provide this information.

[0039] In another implementation example, the tooth model generation neural network also includes a position prediction part configured to output placement coordinates. In this implementation example, the tooth model generation neural network has an additional part that additionally provides the position coordinates. The position prediction part is configured to receive a connection of tooth feature vectors of one or more adjacent teeth as input, and the tooth feature vector is a description of the tooth coordinates of the tooth. For example, the encoder part may include a tooth-specific context encoder neural network for each of the one or more three-dimensional digital tooth models. In this case, each of the tooth-specific context encoder neural networks outputs a tooth feature vector, which is then connected into a collective feature vector. In this case, the position prediction part, for example, connects the two tooth feature vectors of two adjacent teeth of the replacement tooth. Since the adjacent teeth basically determine the position where the replacement tooth can be placed, this is an effective method for determining the position of the replacement tooth.

[0040] In another embodiment, the position prediction portion includes one or more connected layers. This embodiment may be beneficial because the one or more connected layers essentially provide a method of combining tooth coordinates, for example, coordinates of one or more adjacent teeth, to calculate the position coordinates as a linear model. This method of determining the position coordinates is very efficient.

[0041] In another implementation example, the decoder part is implemented as a decoder of an autoencoder.

[0042] In another implementation example, the decoder portion is implemented as a variational autoencoder decoder portion.

[0043] In another implementation example, the decoder portion is implemented as an automatic decoder.

[0044] In another implementation example, the decoder portion is implemented as a generative adversarial network.

[0045] In another implementation example, the decoder part is implemented as a normalized flow model.

[0046] In another implementation example, the decoder part is implemented as a diffusion model.

[0047] In another implementation example, the decoder portion is implemented as an autoregressive model.

[0048] For the decoder part, for each of the embodiments listed above, the context encoder neural network described in the aforementioned implementation examples can also use any one or more neural networks shown in the exemplary. These exemplary neural networks include, for example, any of the following: surface-based neural networks, diffusion network models (DiffusionNet), graph-based neural networks, graph neural networks, point-based neural networks, point network models (PointNet), voxel-based neural networks, three-dimensional shapes (three-dimensional ShapeNet), three-dimensional convolutional neural networks, neural networks based on multi-view rendering, two-dimensional convolutional neural networks, implicit surface based neural networks (implicit surface based neuralnetwork), occupancy networks (Occupancy Network) and decoder parts of convolutional autoencoders. For example, a generative adversarial network can be implemented by using a diffusion network or by using a graph neural network.

[0049] In another implementation example, the latent space vector of the decoder part is regularized. This feature is discussed in the context of variational autoencoders, and it is also applicable to other types of neural networks that can be used as the decoder part. In a normal autoencoder, the encoding part generates a latent space vector that needs to be input to the decoder part later. If the latent space vector is not regularized, the output of the decoder part may be very unstable as its position changes. In essence, when the position of the latent space vector is changed, the vector space will change dramatically in a very short change. In order to obtain a regularized latent space vector, it is necessary to add constraints to the loss function. Usually, the latent space vector will be forced to adapt to a probability distribution, such as a normal distribution. Its effect is the regularization of the latent space. When performing this operation, a slight change in the latent space vector will cause the decoder part to output a slight change. This can be beneficial in this case because it can be achieved that when the latent space vector is slightly changed, the output or generated three-dimensional digital model will change slightly. This makes it possible to improve or optimize.

[0050] In another implementation example, the dental method also includes iteratively modifying the generated three-dimensional digital model of the replacement tooth by modifying the latent space vector using optimization, and modifying the latent space vector by optimizing the loss function that minimizes the loss function used to force interproximal contact with one or more adjacent teeth of the replacement tooth. This is an example application in which the latent space vector of the decoder portion is regularized. Because it has been regularized, an optimization algorithm can be used to modify the generated three-dimensional digital model so that the proximal contact between the replacement tooth and one or more adjacent teeth is correct. For example, this can be achieved by using back propagation and gradient descent methods. However, other optimization algorithms can also be used, and the back propagation is only an example.

[0051] In another implementation example, the loss function further enforces avoiding occlusal contacts greater than a predefined maximum depth. Thus, the occlusal contact of the antagonist tooth of the replacement tooth may be equal to or less than the predefined maximum depth for which a threshold is defined. This implementation example may be beneficial because it may help adjust the generated three-dimensional digital model of the replacement tooth to better fit the subject's jaw. For example, it may also enable the replacement tooth to achieve a suitable occlusion.

[0052] In another implementation example, the loss function further strengthens the contact with the replacement tooth antagonist at the preset mark. This implementation example may be beneficial because it can also help adjust the shape and / or position of the generated three-dimensional digital model of the replacement tooth so that it better fits the subject's jaw. For example, this can enable the replacement tooth to achieve a suitable bite.

[0053] In another implementation example, the method further includes receiving a surface modification. The method further includes applying the surface modification to the generated three-dimensional digital model of the replacement tooth. This implementation example may be beneficial because it may help to make or give the generated three-dimensional digital model a more realistic appearance. The generated three-dimensional digital model is generated using a decoding part controlled by the latent space vector. The latent space vector can accurately provide the correct shape for the generated three-dimensional digital model. However, the dimension of the latent space vector is much lower than that of the actual tooth model, or the dimension of the latent space vector contains much less information than the actual tooth model. Generally speaking, the decoding part, such as the decoding part of the autoencoder, produces a tooth model that is smoother than the actual one. As a result, the teeth generated by the neural network may have a typical artificial appearance. Such tooth models generally lack fine surface details or surface roughness that give the teeth a highly realistic appearance. The surface modification may be a slight modification of the surface to give it the same roughness or texture as the subject's teeth.

[0054] In another implementation example, the surface modification is received from a surface modification database. For example, the surface modification can be retrieved from the database, and in some cases, a dentist or other technician can try using different surface modifications until the generated three-dimensional digital model of the replacement tooth has a more realistic appearance.

[0055] In another implementation example, the method further includes receiving an auto-encoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the subject's selected teeth into a tooth-replicating auto-encoder neural network. The tooth-replicating auto-encoder neural network includes an auto-encoding portion and an auto-encoder decoding portion. The auto-encoder decoding portion is the same as the decoder portion of the tooth-generating neural network. The method further includes determining the surface modification to process the auto-encoded three-dimensional digital tooth model into the three-dimensional digital model of the subject's selected teeth.

[0056] The benefit of this embodiment is that it is able to generate a highly realistic, individual-specific three-dimensional digital model of replacement teeth. In this case, the autoencoder decoding part is the same as the decoder part of the tooth model generation neural network. This means that the latent space vectors of both parts are the same, and the aforementioned smoothing effect from the decoder part is the same in both cases. The selected teeth are taken from the same individual, so the selected teeth have the same surface roughness or topology as the other teeth. Combining the two, the surface structure of the teeth can be copied into the generated three-dimensional digital model of the subject's teeth, which model is not only suitable for the specific decoder part of the tooth model generation neural network, but also for the specific individual.

[0057] For example, the determination of the surface modifications may be performed for a group of subjects and used to compile a database of surface modifications, i.e., a surface modification library. The determined surface modifications may be added to the surface modification library as is. When the determined surface modifications are added to the surface modification library, they may be adjusted, combined, and / or averaged.

[0058] In another embodiment, the selected tooth is a tooth on the same jaw as the replacement tooth. When the replacement tooth is a tooth on the same jaw as the replacement tooth, the method includes mirroring the surface modification. This embodiment is very effective because the tooth on the same jaw as the replacement tooth usually has a very similar appearance and structure to the missing tooth or replacement tooth.

[0059] In another embodiment, the selected tooth is an adjacent tooth to the replacement tooth. This embodiment may be beneficial because teeth adjacent to each other may have very similar surface structures.

[0060] In another implementation example, the selected tooth is an antagonist tooth of the replacement tooth. The antagonist tooth may also have a structure very similar to the replacement tooth, so this implementation may be very effective. The method may include mirroring the surface modification of the antagonist tooth.

[0061] In another implementation example, the decoder portion is configured to output the generated three-dimensional digital model by processing a mesh of a standard shape into the generated three-dimensional digital model of the replacement tooth. In this implementation example, the decoder portion is basically configured to output the generated three-dimensional digital model, and the generated three-dimensional digital model has a specific topological structure of vertices and connections between these vertices. In different examples, the standard shape mesh may take different forms. For example, the standard shape mesh may be a sphere or other shape. The decoder portion is further configured to modify the specific shape into the generated three-dimensional digital model.

[0062] Using the standard shape mesh may be useful when performing style transfer, as changes in one mesh can be applied to another mesh without any problems. For example, the standard shape mesh may be utilized by a tooth replication auto-encoding neural network. The surface modifications used to process the auto-encoded three-dimensional digital tooth model into a three-dimensional digital model may essentially be the coordinates of specific vertices and how they are moved to make the auto-encoded three-dimensional digital tooth model match the selected teeth of the subject. The changes in these coordinates may be the surface modifications. When performing surface modifications on the generated three-dimensional digital model, the changes in the positions of these vertices may simply be vector operations. This approach may be a very effective way to produce realistic teeth through small surface changes in the generated three-dimensional digital model.

[0063] In another implementation example, the method further includes acquiring a three-dimensional digital tooth structure of the subject by using a three-dimensional imaging system. For example, a three-dimensional camera can be used to image the inside of the subject's mouth and used to provide a three-dimensional digital tooth structure. Alternatively, an impression of the subject's tooth structure, i.e., a negative mold of each tooth structure, can be acquired, and the impression can be directly scanned using a three-dimensional camera, or a tooth model can be provided using the impression, i.e., a positive mold of each tooth structure, and the tooth model can be scanned using a three-dimensional camera to provide a three-dimensional digital tooth structure. In other examples, other tomographic imaging techniques, such as computed tomography or even magnetic resonance imaging, can be used to acquire the three-dimensional digital tooth structure. Computed tomography is used more frequently because it is easier for computed tomography to image bone structures.

[0064] In another implementation example, one or more three-dimensional digital tooth models are selected from the segmentation of the three-dimensional digital tooth structure using predetermined criteria. In different examples, the predetermined criteria may be different. In one example, a user or protocol selects one or more three-dimensional digital tooth models based on their spatial relationship with the replacement teeth for which the three-dimensional digital models are to be generated. In another example, the predetermined criteria are determined by the position of the teeth relative to the replacement teeth and based on the content used to train the tooth model generation neural network. For example, specific features of the replacement teeth can be used to select a specific tooth model generation neural network that is trained to replace the specific tooth or a specific type of tooth, such as a molar or other type of tooth. The tooth model generation neural network is trained by using a specific type of input. For example, a specific tooth model that has a spatial relationship with the replacement teeth. This information about which model should be used based on the training input can be used for the predetermined criteria.

[0065] The segmentation of the three-dimensional digital tooth structure may, for example, generate a plurality of three-dimensional digital tooth models, from which one or more three-dimensional digital tooth models are selected for generating a three-dimensional digital model of a replacement tooth according to the predetermined criteria. In addition, the segmentation of the three-dimensional digital tooth structure may, for example, generate a three-dimensional digital model of gums in which teeth described by the plurality of three-dimensional digital tooth models are arranged.

[0066] The segmentation of the three-dimensional digital tooth structure can be performed by using a segmentation module. The segmentation of the three-dimensional digital tooth structure can be performed in a variety of ways. In one example, a shape-deformable segmentation module is used to match a pre-existing elastic tooth model to the three-dimensional digital tooth structure. In another example, the segmentation module can be implemented by a neural network that is trained to receive a three-dimensional digital tooth structure and output a mesh representing individual teeth and gums.

[0067] In another embodiment, the method further comprises rendering a display of a set of one or more three-dimensional digital tooth models of the subject, including the generated three-dimensional digital model of the replacement tooth. In this embodiment, the generated three-dimensional digital model of the replacement tooth is rendered together with other three-dimensional digital tooth models of the subject. For example, this can provide a comparison result between the generated three-dimensional digital model and the other teeth of the subject to show the subject, the dentist, etc. For example, this can be used to approve or disapprove the replacement tooth and prepare the preliminary step of tooth restoration.

[0068] In another implementation example, the method further comprises using the generated three-dimensional digital model of the replacement tooth to make a dental restoration model. The dental restoration may be a crown, an inlay, an onlay or an overlay.

[0069] In another implementation example, the method further includes making a replacement crown model using the generated three-dimensional digital model of the replacement tooth.

[0070] In another implementation example, the method further includes making an inlay model using the generated three-dimensional digital model of the replacement tooth.

[0071] In another implementation example, the method further includes making an onlay model using the generated three-dimensional digital model of the replacement tooth.

[0072] In another implementation example, the method further comprises making an overlay model using the generated three-dimensional digital model of the replacement tooth.

[0073] In all of these embodiments, when making the dental restoration model, such as a replacement crown model, an inlay model, an onlay model or an overlay model, the generated three-dimensional digital model can be used to provide data to a machine to make the replacement or dental restoration. For example, the generated three-dimensional digital model of the replacement tooth can be used as a template for making a dental restoration model, wherein the dental restoration model is a physical copy of at least a portion of the corresponding template. For example, the data can be sent to a manufacturing device, a three-dimensional printing device or a casting device capable of performing this processing step. The casting device can, for example, include a manufacturing device and / or a three-dimensional printing device for making a casting mold, and the casting mold is used, for example, by the casting device to cast the dental restoration.

[0074] In another aspect, the present invention provides a dental system. The dental system includes a memory storing machine executable instructions and a tooth model generation neural network. The tooth model generation neural network includes an encoder portion and a decoder portion. The encoder portion is configured to output a collective feature vector of one or more digital tooth models. The tooth model generation neural network also includes at least one fully connected layer, which is configured to output a latent space vector in response to receiving the collective feature vector. The decoder portion is configured to output the generated three-dimensional digital model of the replacement tooth in response to receiving the latent space vector.

[0075] The dental system also includes a computing system. Execution of the machine executable instructions causes the computing system to receive a selection of a replacement tooth for a subject. Execution of the machine executable instructions also causes the computing system to receive one or more three-dimensional digital tooth models describing one or more teeth of the subject. Execution of the machine executable instructions also causes the computing system to receive a generated three-dimensional digital model of a replacement tooth in response to inputting the one or more three-dimensional digital tooth models describing one or more teeth of the subject into the tooth model generation neural network. The benefits of doing so have been discussed previously.

[0076] The dental system can be configured to perform any of the examples of the aforementioned dental methods.

[0077] In another implementation example, the dental system further comprises a processing system configured to process a dental restoration model using the generated three-dimensional digital model of the replacement tooth. The dental restoration model may be, for example, any of the following: a replacement crown model, an inlay model, an overlay model, and an onlay model.

[0078] The processing system may be, for example, an automated or CNC manufacturing system, a three-dimensional printer, or an automated casting system.

[0079] In some examples, the manufacturing system or the three-dimensional printing system may be used in a process of machining a casting mold.

[0080] In another aspect, the present invention provides a computer program comprising machine executable instructions and a tooth model generation neural network. For example, the computer program may be stored on a non-transitory medium. The tooth model generation neural network comprises an encoder portion and a decoder portion. The encoder portion is configured to output a collective feature vector describing one or more digital tooth models. The tooth model generation neural network also includes at least one fully connected layer, which is configured to output a latent space vector in response to receiving the collective feature vector. The decoder portion is configured to output the generated three-dimensional digital model of the replacement tooth in response to receiving the latent space vector.

[0081] Execution of the machine executable instructions causes the computing system to receive a selection of replacement teeth for a subject. Execution of the machine executable instructions also causes the computing system to receive one or more three-dimensional digital tooth models describing one or more teeth of the subject. Execution of the machine executable instructions also causes the computing system to receive a generated three-dimensional digital model of a replacement tooth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of the subject into the tooth model generation neural network. The benefits of doing so have been discussed previously.

[0082] The computer program may, for example, be configured to perform any of the examples of the aforementioned dental method.

[0083] Another aspect of the present invention is to provide a dental method. The dental method includes receiving a three-dimensional digital model of replacement teeth generated for a subject. The dental method also includes receiving a three-dimensional digital model of selected teeth of the subject. The dental method also includes receiving an auto-encoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the selected teeth of the subject into an auto-encoder neural network. The auto-encoder neural network includes an auto-encoder portion and an auto-encoder decoding portion. The dental method also includes determining a surface modification to deform the auto-encoded three-dimensional digital tooth into the selected teeth of the subject. The dental method also includes applying the surface modification to the generated three-dimensional digital model of replacement teeth.

[0084] The method may be beneficial because it may provide a means to make the generated three-dimensional digital model of replacement teeth of a subject more similar to the subject's actual teeth. A typical problem with the generated digital models of teeth is that they are often smoother than actual or real teeth. For example, the generated digital model of teeth or teeth may lack the texture or fine three-dimensional structure of real teeth.

[0085] In another implementation example, the selected tooth is a tooth on the same jaw as the replacement tooth. If the selected tooth is a tooth on the same jaw as the replacement tooth, surface modification may be performed using mirror transformation before modifying the generated three-dimensional digital model of the replacement tooth. Using the tooth on the same jaw as the replacement tooth may produce a beneficial effect, that is, the tooth on the same jaw as the replacement tooth may determine a suitable surface structure for the replacement tooth based on the symmetry in the human dental arch.

[0086] In another embodiment, the selected tooth is an adjacent tooth of the replacement tooth. Using the adjacent tooth of the replacement tooth may have a beneficial effect, that is, the adjacent tooth can determine a suitable surface structure for the replacement tooth. This is especially true if the adjacent tooth is of the same type as the replacement tooth.

[0087] In another implementation example, the selected tooth is an antagonist tooth of the replacement tooth. If the selected tooth is an antagonist tooth, in some examples, the dental method may further include performing surface modification using a mirror transformation before modifying the generated three-dimensional digital model of the replacement tooth. Using an antagonist tooth of the replacement tooth may have a beneficial effect, that is, due to the symmetry between the two dental arches of humans, the antagonist tooth may determine a suitable surface structure for the replacement tooth.

[0088] The dental method also includes receiving a selection of replacement teeth of a subject. The dental method also includes receiving one or more three-dimensional digital tooth models describing one or more teeth of the subject. The dental method also includes receiving the generated three-dimensional digital model of the replacement teeth of the subject in response to inputting the one or more three-dimensional digital tooth models into the tooth model generation neural network. The tooth model generation neural network includes an encoder part and a decoder part. The decoder part is the same as the automatic encoder decoding part. The encoder part is configured to output a collective feature vector describing one or more teeth of the subject. The generation neural network also includes at least one fully connected layer, which is configured to output a latent space vector in response to receiving the collective feature vector. The decoder part is configured to output the three-dimensional digital model in response to receiving the latent space vector.

[0089] In another implementation example, the generated three-dimensional digital model of the replacement tooth is received from a database of generated three-dimensional digital tooth models, such as a tooth library. In the above implementation example, the generated three-dimensional digital model is generated by a tooth model generation neural network. This implementation example is an alternative. For example, there may be a large model library or database containing a large number of generated three-dimensional digital models. For example, one or some of the models may be suitable as replacement teeth or replacement tooth models for a subject, but they may lack the same surface structure and fine details of the subject's real teeth. Such differences in surface structure and / or details may be caused, for example, by individual wear of the subject's real teeth. This implementation example provides a method of using a model library tooth as a replacement tooth and then modifying the model library tooth to have the appearance of a subject's natural teeth.

[0090] In another implementation example, the decoder portion is configured to generate the generated three-dimensional digital model of the replacement tooth by processing a standard shape mesh. Using a standard shape mesh may be beneficial because, as described above, it provides a standard number of vertices and the relationship between these vertices. Therefore, the first surface modification may be just instructions on how to translate the individual vertices of the standard shape mesh. It can provide a standardized method for performing surface modifications on the three-dimensional digital model of the tooth.

[0091] In another embodiment, the standard shape mesh includes vertices. The surface modification is determined by calculating the translation of the vertices of the three-dimensional digital model of the selected tooth along the normal until they intersect with the automatically encoded three-dimensional digital tooth model. For example, this can provide a standardized method for modifying a standard shape mesh. For example, the normals are vertex normals, i.e., direction vectors associated with the vertices, intended to replace the true geometric normals of the surface defined by the mesh. These vertex normals can be calculated as the normalized average of the surface normals of the surface defined by the mesh containing the corresponding vertices. For example, the average can be weighted according to the surface area or can be unweighted. In order to determine the intersection and intersection point with the automatically encoded three-dimensional digital tooth model, the automatically encoded three-dimensional digital tooth model can be superimposed on the three-dimensional digital model of the selected tooth.

[0092] In another implementation example, the surface modification is determined by displacement mapping between the three-dimensional digital model of the selected tooth and the automatically encoded three-dimensional digital tooth model. The displacement mapping used here represents the displacement of the tooth surface. Providing a displacement mapping may also be an effective method for adding fine structure or texture to the three-dimensional digital tooth model.

[0093] In another aspect, the present invention provides a dental system. The dental system includes a memory storing machine executable instructions. The dental system also includes a computing system. Execution of the machine executable instructions causes the computing system to receive the generated three-dimensional digital model of the replacement teeth of the subject. Execution of the machine executable instructions also causes the computing system to receive the generated three-dimensional digital model of the replacement teeth of the subject. Execution of the machine executable instructions also causes the computing system to receive the autoencoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the selected teeth of the subject into the autoencoder neural network. The autoencoder neural network includes an autoencoder encoding portion and an autoencoder decoding portion.

[0094] Execution of the machine executable instructions further causes the computing system to determine a surface modification to process the automatically encoded three-dimensional digital tooth as a selected tooth of a subject. Execution of the machine executable instructions further causes the computing system to apply the surface modification to the generated three-dimensional digital model of the replacement tooth.

[0095] The dental system can be configured to perform any of the examples of the aforementioned dental methods.

[0096] In another implementation example, the dental system further comprises a processing system configured to process a dental restoration model using a generated three-dimensional digital model of a replacement tooth, the generated three-dimensional digital model having surface modification applied thereto. The dental restoration model may be, for example, any of the following: a replacement crown model, an inlay model, an overlay model, and an onlay model.

[0097] Another aspect of the present invention is to provide a computer program including machine executable instructions. For example, the computer program can be stored on a non-temporary storage medium. Execution of the machine executable instructions causes the computing system to receive a generated three-dimensional digital model of a replacement tooth of a subject. Execution of the machine executable instructions further causes the computing system to receive a three-dimensional digital model of selected teeth of the subject. Execution of the machine executable instructions also causes the computing system to receive an auto-encoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the selected teeth of the subject into an auto-encoder neural network. The auto-encoder neural network includes an auto-encoding portion and an auto-encoder decoding portion.

[0098] Execution of the machine executable instructions also causes the computing system to determine a surface modification to process the automatically encoded three-dimensional digital tooth as a selected tooth of a subject. Execution of the machine executable instructions further causes the computing system to apply the surface modification to the generated three-dimensional digital model of the replacement tooth.

[0099] The computer program may be configured to perform any of the examples of the aforementioned dental method.

[0100] Another aspect of the present invention is to provide a dental method. The method includes receiving a generated three-dimensional digital model of a subject's replacement teeth. The generated three-dimensional digital model is formed by a topological mesh having vertices. The method also includes receiving a surface modification, which is to be transferred to the mesh forming the generated three-dimensional digital model. The surface modification is for the same mesh topology as the mesh forming the generated three-dimensional digital model, and includes position adjustment of the mesh topology vertices. The method also includes applying the surface modification to the mesh forming the generated three-dimensional digital model of the replacement teeth. In this implementation example, this may mean using the position adjustment to change the position of the vertices of the generated three-dimensional digital model of the subject's replacement teeth.

[0101] The method may be beneficial because it may provide an advantageous means of making a three dimensional digital model of replacement teeth generated for a subject more similar to the subject's actual teeth by transferring surface modifications to the mesh forming the generated three dimensional digital model of the replacement teeth. The surface modifications define positional adjustments to mesh topology vertices of the mesh forming the generated three dimensional digital model of the replacement teeth. A typical problem with generated digital models of teeth is that they may be smoother than actual or real teeth, or that such models may not have fine structure that matches the subject's other teeth. For example, teeth may lack small details, such as texture or fine three dimensional structure, or they may differ in these small details.

[0102] The received surface modifications to be transferred to the mesh forming the generated three-dimensional digital model may be surface modifications defined for a specific mesh topology of vertices and connections between these vertices, which mesh topology is the same as the mesh topology of the mesh forming the generated three-dimensional digital model of the replacement tooth. Thus, modifications defined by one mesh, such as surface modifications, can be easily applied to meshes with the same topology. The surface modifications may essentially be the coordinates of specific vertices and how they are moved. When performing surface modifications on the generated three-dimensional digital model of the replacement tooth, changes in the positions of these vertices may simply be applied as vector operations. This may be a very effective way to produce realistic teeth by small surface changes in the generated three-dimensional digital model of the replacement tooth.

[0103] In another example implementation, the surface modification is received from a surface modification database. In this example implementation, a library of three-dimensional digital models of teeth may be used and one of the models may be selected as the model for the replacement tooth. However, its surface may not match the surface of the subject's teeth. Using the surface modification may provide an improved method for providing a subject with a replacement tooth that looks realistic.

[0104] In another implementation example, the method further includes receiving an auto-encoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the subject's selected teeth into a tooth replication autoencoder neural network. The tooth replication autoencoder neural network includes an autoencoder encoding portion and an autoencoder decoding portion. The autoencoder decoding portion is the same as the decoder portion of the tooth model generation neural network. The method further includes determining a surface modification to process the auto-encoded three-dimensional digital tooth model into a three-dimensional digital model of the subject's selected teeth.

[0105] In another implementation example, the selected tooth is a tooth on the opposite side of the same jaw. If the selected tooth is a tooth on the opposite side of the same jaw, the method may further include modifying the surface using a mirror transformation. In another implementation example, the selected tooth is an adjacent tooth of the replacement tooth.

[0106] In another implementation example, the selected tooth is an antagonist tooth of the replacement tooth. If the selected tooth is an antagonist tooth, the method may further include modifying the surface using a mirror transformation.

[0107] Another aspect of the present invention is to provide a dental system. The dental system includes a memory storing machine executable instructions. The dental system also includes a computing system. Execution of the machine executable instructions causes the computing system to receive a generated three-dimensional digital model of a replacement tooth of a subject. The generated three-dimensional digital model is formed by a mesh having a mesh topology of vertices. Execution of the machine executable instructions further causes the computing system to receive a surface modification to be transferred to the mesh forming the generated three-dimensional digital model. The surface modification is defined for the same mesh topology as the mesh forming the generated three-dimensional digital model and includes positional adjustments to the mesh topology vertices. Execution of the machine executable instructions further causes the computing system to apply the surface modification to the mesh forming the generated three-dimensional digital model of the replacement tooth.

[0108] The dental system can be configured to perform any of the examples of the aforementioned dental methods.

[0109] In another implementation example, the dental system further comprises a processing system configured to manufacture a dental restoration model using the generated three-dimensional digital model of the replacement tooth. For example, the processing system may include a CNC or automated manufacturing system, a three-dimensional printer, or a system for preparing a casting. The dental restoration model may be, for example, any of the following: a replacement crown model, an inlay model, an overlay model, and an onlay model.

[0110] Another aspect of the present invention is to provide a computer program including machine executable instructions. For example, the computer program can be stored on a non-transitory storage medium. Execution of the machine executable instructions causes the computing system to receive a generated three-dimensional digital model of a replacement tooth of a subject. The generated three-dimensional digital model is formed by a mesh having a mesh topology of vertices. Execution of the machine executable instructions further causes the computing system to receive a surface modification to be transferred to the mesh forming the generated three-dimensional digital model. The surface modification is defined for the same mesh topology as the mesh forming the generated three-dimensional digital model and includes positional adjustments to the mesh topology vertices. Execution of the machine executable instructions further causes the computing system to apply the surface modification to the mesh forming the generated three-dimensional digital model of the replacement tooth.

[0111] The computer program may be configured to perform any of the examples of the dental method described above.

[0112] It should be understood that one or more of the aforementioned implementation examples of the present invention may be combined as long as the combined implementation examples are not mutually exclusive.

[0113] It will be appreciated by those skilled in the art that various aspects of the present invention may be embodied as an apparatus, method or computer program product. Therefore, various aspects of the present invention may take the form of a completely hardware implementation example, a completely software implementation example (including firmware, resident software, microcode, etc.) or an implementation example combining software and hardware aspects, which may generally be referred to herein as a "circuit", "module" or "system". In addition, various aspects of the present invention may take the form of a computer program product embodied in one or more computer-readable media having computer executable code thereon.

[0114] Any combination of one or more computer-readable media may be utilized. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. As used herein, "computer-readable storage medium" includes any tangible storage medium that can store instructions that can be executed by a processor or a computing system of a computing device. The computer-readable storage medium may refer to a computer-readable non-temporary storage medium. The computer-readable storage medium may also refer to a tangible computer-readable medium. In some implementation examples, the computer-readable storage medium may also store data that can be accessed by a computing system of a computing device. Examples of the computer-readable storage medium include, but are not limited to, a floppy disk, a magnetic hard disk drive, a solid-state hard disk, a flash memory, a USB thumb drive, a random access memory (RAM), a read-only memory (ROM), an optical disk, a magneto-optical disk, and a register file of a computing system. Examples of optical disks include compact disks (CDs) and digital versatile disks (DVDs), such as CD-ROMs, CD-RWs, CD-Rs, DVD-ROMs, DVD-RWs, or DVD-R disks. The term "computer-readable storage medium" also refers to various types of recording media that can be accessed by a computer device via a network or communication link. For example, data can be retrieved via a modem, the Internet, or a local area network. Computer executable code embodied on a computer readable medium may be transmitted via any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, radio frequency (RF), etc., or any suitable combination of the foregoing.

[0115] A computer readable signal medium may include a propagated data signal with computer executable code embedded therein, for example, in baseband form or as part of a carrier wave. Such propagated signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium, and may communicate, propagate, or transport a program for use with an instruction execution system, apparatus, or device, and combinations thereof.

[0116] "Computer internal memory" or "internal memory" is an example of a computer-readable storage medium. Computer memory is any memory directly accessible by a computing system. "Computer storage" or "memory" is another example of a computer-readable storage medium. Computer memory is any non-volatile computer-readable storage medium. In some implementation examples, computer memory may also be computer internal memory and vice versa.

[0117] As used herein, a "computing system" includes an electronic component capable of executing a program or executing a machine-executable instruction or executing a computer-executable code. References to computing systems including examples of "computing systems" should be interpreted as possibly including more than one computing system or processing core. The computing system may be, for example, a multi-core processor. A computing system may also refer to a collection of computing systems within a single computer system or distributed among multiple computer systems. The term "computing system" should also be interpreted as possibly referring to a collection or network of computing devices, each of which includes a processor or computing system. Machine executable code or instructions may be executed by multiple computing systems or processors located within the same computing device or even distributed across multiple computing devices.

[0118] Machine executable instructions or computer executable code may include instructions or programs that enable a processor or other computing system to perform an aspect of the present invention. Computer executable code for performing the operations of various aspects of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages, such as Java, Smalltalk, C++, etc., and traditional process programming languages, such as "C" programming language or similar programming languages, and compiled into machine executable instructions. In some cases, computer executable code may be in the form of a high-level language or in a precompiled form, and used together with a machine executable instruction interpreter that can be generated instantly. In other examples, machine executable instructions or computer executable code may be the programming form of a programmable logic gate array.

[0119] The computer executable code may execute entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet through an Internet service provider).

[0120] Aspects of the present invention are described with reference to the flow chart and / or block diagram of the method, device (system) and computer program product according to the implementation example of the present invention. It should be understood that each block or partial block in the flow chart and / or block diagram can be implemented by the computer program instruction in the form of computer executable code when applicable. It should also be understood that when not mutually exclusive, the blocks in different flow charts and / or block diagrams can be combined. These computer program instructions can be provided to the computing system of a general-purpose computer, a special-purpose computer or other programmable data processing device so that the instruction executed by the computing system of the computer or other programmable data processing device is created to realize the function / action specified in the flow chart and / or block diagram block.

[0121] These machine-executable instructions or computer program instructions may also be stored in a computer-readable medium, which may instruct a computer, other programmable data processing device, or other device to operate in a specific manner so as to execute the instructions stored in the computer-readable medium, including instructions for executing the functions / actions specified in the flowchart and / or block diagram blocks to achieve the purposes of the present invention.

[0122] Machine executable instructions or computer program instructions may also be loaded onto a computer or other programmable data processing device or other device so that a series of operational steps are performed on the computer, other programmable device or other device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide a process for implementing the functions / actions in one or more functional blocks specified in the flowchart and / or block diagram.

[0123] As used herein, a "user interface" is an interface that allows a user or operator to interact with a computer or computer system.

[0124] A "user interface" may also be referred to as a "human-computer interaction device". A user interface may provide information or data to an operator and / or receive information or data from an operator. A user interface may enable a computer to receive input from an operator and to provide output from the computer to a user. In other words, a user interface may allow an operator to control or manipulate a computer and the interface may allow the computer to indicate the effects of the operator's control or manipulation. Displaying data or information on a display or graphical user interface is an example of providing information to an operator. Receiving data through a keyboard, mouse, trackball, touchpad, pointing stick, graphics tablet, joystick, game controller, webcam, headset, pedal, wired gloves, remote control, and accelerometer are all examples of user interface components that may receive information or data from an operator.

[0125] As used herein, a "hardware interface" includes an interface that enables a computing system of a computer system to interact with and / or control an external computing device and / or apparatus. A hardware interface may allow a computing system to send control signals or instructions to an external computing device and / or apparatus. A hardware interface may also enable a computing system to exchange data with an external computing device and / or apparatus. Examples of hardware interfaces include, but are not limited to, a universal serial bus, an IEEE 1394 port, a parallel port, an IEEE 1284 port, a serial port, an RS-232 port, an IEEE-488 port, a Bluetooth connection, a wireless LAN connection, a TCP / IP connection, an Ethernet connection, a control voltage interface, a MIDI interface, an analog input interface, and a digital input interface.

[0126] As used herein, a "display" or "display device" includes an output device or user interface suitable for displaying images or data. A display may output visual, audio, or tactile data. Examples of displays include, but are not limited to, computer monitors, television screens, touch screens, tactile electronic displays, Braille screens, cathode ray tubes (CRTs), storage tubes, bi-stable displays, electronic paper, vector displays, flat panel displays, vacuum fluorescent displays (VFs), light emitting diode (LED) displays, electroluminescent displays (ELDs), plasma display panels (PDPs), liquid crystal displays (LCDs), organic light emitting diode displays (OLEDs), projectors, and head mounted displays. BRIEF DESCRIPTION OF THE DRAWINGS

[0127] The following is a more detailed explanation of the implementation examples of the present invention by way of examples and with reference to the accompanying drawings, in which

[0128] Figure 1 An example implementation of a dental system is shown;

[0129] Figure 2 Shows the instructions for operation Figure 1 A flow chart of the method of the dental system;

[0130] Figure 3 Another implementation example of a dental system is shown;

[0131] Figure 4 Shows the instructions for operation Figure 3 A flow chart of the method of the dental system;

[0132] Figure 5 Shows the instructions for operation Figure 3 A flow chart of the method of the dental system;

[0133] Figure 6 An implementation example of a three-dimensional digital tooth structure is shown;

[0134] Figure 7 An example of segmentation of a digital tooth structure is shown;

[0135] Figure 8 The structure of the neural network for generating tooth models is shown;

[0136] Fig. 9 A method of transferring natural tooth patterns to teeth generated by a neural network using an autoencoder is shown;

[0137] Fig.10 Shows Fig. 9 The results of the method;

[0138] Fig.11 Another implementation example of a dental system is shown; and

[0139] Fig.12 Another implementation example of a dental system is shown. DETAILED DESCRIPTION

[0140] Elements with the same number in the above-mentioned implementation example figures are equivalent elements or elements that perform the same function. If the functions are equivalent, the elements discussed previously may not be discussed in the following figures.

[0141] Figure 1 An example implementation of a dental system 100 is shown. The dental system 100 described in this example implementation includes a computer 102, an optional digital model source 104, and an optional processing system 106. The computer 102 also includes an optional display 108. The digital model source 104 can be, for example, a model library that provides digital models, or in other examples, the digital model source 104 can be an optical or other medical scanner for obtaining a measured three-dimensional model from a subject.

[0142] The computer 102 is intended to represent one or more computing systems or computers located at one or more locations. The computer 102 includes a computing system 110. Likewise, the computing system 110 is intended to represent one or more computing systems or processing cores located at one or more locations. The computing system 110 is in communication with an optional hardware interface 112 and an optional user interface 114. The hardware interface 112 enables the computing system 110 to control and communicate with other components of the dental system 100. In some cases, this is a dedicated hardware interface 112, or it may be a network interface that allows the computing system 110 to communicate remotely with other components via a network or Internet connection.

[0143] The user interface 114 enables an operator or user of the dental system 100 to interact with and / or control the dental system 100. In this implementation example, the user interface 114 optionally includes a display 108.

[0144] The computing system 110 is in communication with a memory 116. The memory 116 may be any type of memory accessible to the computing system 110. In some implementation examples, the memory 116 is a non-transitory storage medium.

[0145] The memory 116 includes machine executable instructions 120. The machine executable instructions 120 enable the computing system 110 to perform various data processing and numerical tasks and control other components of the dental system 100. The memory 116 further includes a tooth model generation neural network 122. The memory 116 further includes a selection of replacement teeth 124. The memory 116 further includes one or more three-dimensional digital tooth models 126. For example, the one or more three-dimensional digital tooth models 126 may be received from the digital model source 104. The memory 116 also includes a generated three-dimensional digital model of the replacement tooth received in response to inputting the one or more three-dimensional digital tooth models 126 into the tooth model generation neural network 122.

[0146] The tooth model generating neural network 122 can be configured or trained in several different ways. In one case, the tooth model generating neural network 122 can be trained so that it can take a variety of one or more three-dimensional digital tooth models 126 as input and then output different types of teeth. Another option is that there is an optional set of tooth model generating neural networks 130 for each specific tooth or tooth type. For a specific tooth or tooth type, the generated three-dimensional digital model 128 of the tooth can be selected to replace.

[0147] In some cases, the generated three-dimensional digital model of replacement teeth 128 is configured to optionally also output position coordinates of the generated three-dimensional digital model of replacement teeth. The memory 116 optionally contains the position coordinates 132.

[0148] In some examples, the tooth model generating neural network 122 may have a regularized latent space vector. In this case, the latent space vector may be continuously modified, and the generated three-dimensional digital model of replacement teeth 128 may be continuously modified by modifying the latent space vector. The memory 116 also includes an optional optimization algorithm 134, which can be used to control such modification of the latent space vector. For example, this can be used to modify the replacement teeth so that the replacement teeth can better fit the space between adjacent teeth and the opposing teeth.

[0149] The memory 116 also optionally includes surface modifications. When the three-dimensional digital model 128 of the replacement tooth is generated using the neural network, it is generally smoother than a real tooth. The surface modification may be a modification used to texturize and add structure to the surface so that the generated three-dimensional digital model 128 of the replacement tooth has a more realistic or natural appearance. The memory 116 also includes a three-dimensional digital model 138 of the replacement tooth generated after the optional surface modification is applied. This is the three-dimensional digital model 128 of the replacement tooth generated after the surface is modified or textured using the surface modification 136.

[0150] After the generated three-dimensional digital model 128 of the replacement tooth is provided, it can be used for a variety of different purposes. The memory 116 also optionally includes a computer-aided design module 140. The computer-aided design module 140 takes the generated three-dimensional digital model 128 of the replacement tooth as input and uses it to generate control commands 142 for the machining system 106. The machining system 106 can then be used to machine a dental restoration model, such as a replacement crown model, an inlay model, an overlay model, or an onlay model. For example, the machining system can be a computer-controlled manufacturing system, a three-dimensional printer, or a casting system.

[0151] The display 108 may optionally be a rendering of a set of one or more three-dimensional digital tooth models 150 together with the generated three-dimensional digital model of the replacement teeth 128. Such rendering of the display 108 may be beneficial because an operator or even a dentist may see the generated three-dimensional digital model of the replacement teeth 128 in the context of the subject's other teeth.

[0152] Figure 2 Shows the instructions for operation Figure 1 Flowchart of the method of the dental system 100. First, in step 200, a selection 124 of the replacement teeth is received. Next, in step 202, one or more three-dimensional digital tooth models 126 are received. After step 202, step 204 is optionally performed. In step 204, a tooth model generating neural network 122 is selected from the set 130 of tooth model generating neural networks. After this step 204, in step 206, a generated three-dimensional digital model of the replacement teeth 128 is received in response to inputting the one or more three-dimensional digital tooth models 126 into the tooth model generating neural network 122. After step 206, step 208 is optionally performed.

[0153] In step 208, the generated three-dimensional digital model 128 of the replacement tooth is iteratively modified by modifying the latent space vector using an optimization minimization loss function. For example, this can be performed using the optimization algorithm 134. Applicable restrictions have been discussed above. Next, step 210 is optionally performed. In step 210, the position coordinates 132 are output from the tooth model generation neural network 122. After this step, step 212 is optionally performed. In step 212, a surface modification 136 is received. Thereafter, step 214 is optionally performed. In step 214, the surface modification 136 is applied to the generated three-dimensional digital model 128 of the replacement tooth to produce a surface-modified generated three-dimensional digital model 138 of the replacement tooth. Thereafter, step 216 is optionally performed.

[0154] In step 216, a set of one or more three-dimensional digital tooth models 150 of the subject are rendered together with the generated three-dimensional digital model 128 of the replacement teeth. After performing step 216, step 218 is optionally performed. In step 218, the generated three-dimensional digital model of the replacement teeth can be used to manufacture a dental restoration, such as a crown or other implant or onlay.

[0155] Figure 3 Another implementation example of the dental system 300 is shown. As shown, the dental system includes a computer 102 and optionally a digital model source 104 and a processing system 106. In this implementation example, the memory 116 contains machine executable instructions 120. Again, these machine executable instructions 120 enable the computing system 110 to perform various computational and numerical tasks and possible control of the digital model source 104 and the processing system 106.

[0156] The memory 116 also contains the generated three-dimensional digital model 128 of the replacement tooth and the surface modifications 136 and the surface modified generated three-dimensional digital model 138 of the replacement tooth. The memory 116 also contains the three-dimensional digital model 302 of the selected tooth. The memory also contains the autoencoder neural network 304. The memory also contains the autoencoded three-dimensional digital tooth model 306, which is created by inputting the three-dimensional digital model 302 of the selected tooth into the autoencoder neural network 304. As an alternative to using the digital model source 104, a three-dimensional digital tooth model library 308 can be used. The memory also optionally contains a surface modification database 310. One method of surface modification is to save the surface modifications after they are created so that they can be reapplied to other teeth.

[0157] Figure 4 Shows the use of Figure 3The method of the dental system 300 is described. First, in step 400, the generated three-dimensional digital model 128 of the replacement teeth is received. This can be received, for example, from the digital model source 104 or from a three-dimensional digital tooth model library. Next, in step 402, the three-dimensional digital model 302 of the selected teeth is received. Next, in step 404, the automatically encoded three-dimensional digital tooth model 306 is received in response to inputting the three-dimensional digital model 302 of the selected teeth into the autoencoder neural network 304. Next, in step 406, the surface modification 136 is determined, which can deform the automatically encoded three-dimensional digital tooth model 306 into the three-dimensional digital model of the selected teeth. Finally, in step 214, the surface modification 136 is applied to the generated three-dimensional digital model 128 of the replacement teeth.

[0158] Figure 5 Another method of using the dental system 300 is shown. First, in step 400, the generated three-dimensional digital model 128 of the replacement tooth is received, which can come from the three-dimensional digital tooth model library 308 or from the digital model source 104. Next, in step 212, the surface modification 136 is received, which can be as follows: Figure 4 The surface modifications 136 are then received as shown, or may be received from a surface modification database 310 located in the memory 116. The surface modifications 136 are then applied to the generated three-dimensional digital model 128 of the replacement tooth at step 214.

[0159] It is worth noting that Figure 4 and / or Figure 5 The method shown can be used with Figure 2 Similarly, Figure 3 The characteristics can also be Figure 1 feature combination.

[0160] Figure 6 An implementation example of a three-dimensional digital tooth structure 600 is shown. The three-dimensional digital tooth structure 600 described in this implementation example is a three-dimensional model of a subject's teeth obtained by an optical scanner. In the figure, a segmentation section 602 is located next to the three-dimensional digital tooth structure 600. The segmentation section 602 includes a three-dimensional digital tooth model 604 and gums 606. It can be seen that the three-dimensional digital tooth model 604 provides a separate model or a three-dimensional model for each segmented tooth.

[0161] Figure 7 A further segmentation 602 is shown, but in this case there is a missing tooth 700. In the illustration, the right side of the segmentation 602 shows an example of filling the position of the missing tooth 700 using the generated three-dimensional digital model 128 of the replacement tooth.

[0162] Figure 8 The structure and operation of the tooth model generating neural network 122 are shown. It includes an encoder part 804 and a decoder part 806. A plurality of separate three-dimensional digital tooth models 126 are input into the encoder part 804. In this case, it is the central neighbor tooth 808, the distal neighbor tooth 810 and the contralateral tooth 812 of the same jaw. Above the diagram of the tooth model generating neural network 122 is an example of a mesh model for the input mesh 800 and the output mesh 802. The input in the figure shows the position of the central neighbor tooth 808, the distal neighbor tooth 810 and the contralateral tooth 812 of the same jaw and the missing tooth 700. The output shows the same situation, except that the generated three-dimensional digital model 128 of the replacement tooth is inserted at the position of the missing tooth point 700.

[0163] The encoder portion 804 is shown as having a tooth-specific context encoder 814 that individually receives each of the one or more three-dimensional digital tooth models 126. Each tooth-specific context encoder 814 then outputs its tooth feature vector 816. The tooth feature vectors 816 are then concatenated into a collective feature vector 818. The collective feature vector 818 then passes through several fully connected layers 820, where the collective feature vector 818 is converted into a latent space vector 822. The latent space vector 822 then controls the decoder portion 806 to output the generated three-dimensional digital model 128 of the replacement tooth. In this example, the decoder portion 806 takes a canonical shape 824 and warps it into the generated three-dimensional digital model 128 of the replacement tooth.

[0164] The tooth model generation neural network 122 may also optionally include a position prediction section 826, which inputs the additional feature vector 827 to a fully connected layer 828 and then outputs a position coordinate 830. The position coordinate 830 may be provided separately or may be appended to the generated three-dimensional digital model 128 of the replacement teeth to provide the generated three-dimensional digital model 128' of the replacement teeth with the position coordinates. For example, the position coordinates may be simply appended to it, or the position coordinates may be used to move the actual position of the three-dimensional digital model.

[0165] Below the diagram of the tooth model generating neural network 122 is a diagram of the autoencoder neural network 840. It is used to illustrate how the decoder portion 806 is trained before the rest of the tooth model generating neural network 122. The autoencoder neural network 840 includes an autoencoder encoding portion 842 and an autoencoder decoding portion 844. There is a latent space vector 822 between them. This can be trained by inputting different three-dimensional digital models of selected teeth as input and then comparing them to the autoencoded three-dimensional digital tooth model 848. A potential benefit of training the decoder portion 806 separately is that less training data is required to more accurately generate the generated three-dimensional digital model 128 of the replacement teeth. Another benefit is that the autoencoder neural network 840 used in this training can be reused to perform style transfer between the subject's teeth and the generated three-dimensional digital model of the replacement teeth. This may make the replacement teeth look more realistic than the effect obtained by using only the neural network.

[0166] Fig. 9 It is shown how to use an autoencoder neural network 840 for training to provide style transfer between a subject's existing teeth and teeth generated by the neural network. The autoencoder neural network 840 takes as input a three-dimensional digital model 846 of selected teeth of the subject. For example, it can be a three-dimensional digital model of teeth obtained using an optical scanner. The three-dimensional digital model 846 of the selected teeth shows the detailed structure of the subject's teeth. It is then passed through the autoencoder neural network to obtain an autoencoded three-dimensional digital tooth model 848. The teeth of the autoencoded three-dimensional digital tooth model 848 are similar to the three-dimensional digital model 846 of the selected teeth, but have been artificially smoothed.

[0167] The difference between the three-dimensional digital model 846 of the selected tooth and the automatically encoded three-dimensional digital tooth model 848 tooth is calculated by the normals along the planes defined by the vertices. This produces a series of vectors that can be used to translate these surfaces. These vectors are an example of a surface modification 136. Below this, the decoder portion 806 of the tooth model generation neural network 122 receives the generated three-dimensional digital model 128 of the replacement tooth. The surface modification 136 is then applied to the same normals to obtain the surface modified generated three-dimensional digital model 138 of the replacement tooth. Since the decoder portion 806 is the same as the autoencoder decoding portion 844, it smoothes the model in the same way. Therefore, the surface modification 136 is able to add the same type of surface roughness and structure to make the surface modified generated three-dimensional digital model of the replacement tooth look more realistic.

[0168] Fig.10This is illustrated. The tooth labeled 128 is the generated three-dimensional digital model of the replacement tooth. The tooth labeled 846 is the three-dimensional digital model of the selected tooth, which is the tooth selected to provide the style that is subsequently added to the generated three-dimensional digital model of the replacement tooth 128. The tooth labeled 138 is the modified tooth 128 with the style of the tooth 846 added. It can be seen that the tooth 138 looks more realistic and more like the style of the tooth 846. The tooth 128, on the other hand, appears to have been artificially smoothed.

[0169] Fig.11 An example implementation of a dental system 100 having a computer device 102 is shown, similar to Figure 1 or Figure 3 The computer device 102 is configured to execute Figure 2 , Figure 4 and / or Figure 5 The computer device 102 may include a hardware component 54, wherein the hardware component 54 includes one or more processors and a memory storing machine executable program instructions. The one or more processors execute the program instructions so that the one or more processors control the computer device 102 to perform Figure 2 , Figure 4 and / or Figure 5 One or more of the dental methods described herein.

[0170] The computer device 102 may also include one or more input devices, such as a keyboard 58 and a mouse 56, to enable a user to interact with the computer device 102. Furthermore, the computer device 102 has one or more output devices, such as a display 108, which provides a graphical user interface 50 having control elements 52 (e.g., GUI elements) to enable a user to control the execution of the computer device 102. Figure 2 , Figure 4 and / or Figure 5 The display 108 displays the generated three-dimensional digital model 128 of the replacement tooth. The computer device 102 also includes a three-dimensional imaging system in the form of an optical scanner 59, which is configured to acquire a three-dimensional digital tooth structure of the subject. The scanner 59 may be, for example, an intraoral scanner, which is configured to scan the tooth structure in the mouth of the subject. The scanner 59 may, for example, be a scanner configured to scan an impression of the tooth structure of the subject (i.e., a negative mold of the corresponding tooth structure) and / or for scanning a used dental mold (i.e., a positive mold of the corresponding tooth structure).

[0171] Fig.12 Shows Fig.11The dental system 100 further comprises an exemplary machining system 106 for machining a physical dental restoration model 80 using the generated three-dimensional digital model 128 of the replacement tooth as a template. The dental restoration model 80 being made is a physical copy of at least a portion of the corresponding three-dimensional digital model 128. The computer device 102 may be configured to control one or more machining devices 60, 70. For example, the machining system 106 may include a machining device in the form of a manufacturing device 70 controlled by the computer device 102. The manufacturing device 70 may be configured to machine a blank 76 using one or more manufacturing tools 72. A holding device 74 may be used to fix the blank 76 of raw material 78 and cut it into the desired shape and size of the element to be manufactured, for example, a dental restoration model 80. The manufacturing tool 72 may be, for example, a milling tool.

[0172] For example, the processing system 106 may include a processing device in the form of a three-dimensional printing device 60. The three-dimensional printing device 60 may be controlled by the computer device 102 and configured to print the element to be manufactured, such as a dental restoration model 80. The three-dimensional printing device 60 may include a printing element 62, which is configured to print the corresponding element, such as the dental restoration model 80, layer by layer. The printing element 62 may, for example, include a nozzle configured to dispense a printing material.

[0173] For example, one or more processing devices 60, 70 may be configured to process a casting mold, which is then injected with casting material by an additional casting device (not shown) to cast a dental restoration model 80. The casting material is then cured to form a dental restoration model 80.

[0174] Although the present invention has been described with reference to specific implementation examples, it should be understood that the present invention is not limited to the above-mentioned implementation examples, and that those skilled in the art can easily think of many variations of these implementation examples after reading the present disclosure. Therefore, the present invention can be further described by the following implementation examples, but is not limited thereto and is only used as an example. The following implementation examples may include preferred implementation examples. Therefore, the term "clause" used may refer to "preferred implementation examples". Clause 1: A dental method, wherein the method comprises:

[0175] - Receive the subject's selection of replacement teeth;

[0176] - receiving one or more three-dimensional digital tooth models describing one or more teeth of the subject; and

[0177] - Receiving a generated three-dimensional digital model of replacement teeth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of a subject into a tooth model generating neural network, wherein the tooth model generating neural network includes an encoder part and a decoder part, wherein the encoder part is configured to output a collective feature vector describing the one or more digital tooth models, wherein the tooth model generating neural network also includes at least one fully connected layer, which is configured to output a latent space vector in response to receiving the collective feature vector, and wherein the decoder part is configured to output the generated three-dimensional digital model of replacement teeth in response to receiving the latent space vector.

[0178] Clause 2: The dental method of clause 1, wherein the method further comprises using the selection of replacement teeth to select a tooth model generation neural network for generating replacement teeth from a set of tooth model generation neural networks trained for specific teeth or specific tooth types.

[0179] Item 3: A dental method according to Item 1 or 2, wherein the encoder portion includes a tooth-specific context encoder neural network for each of one or more three-dimensional digital tooth models, wherein the tooth-specific context encoder neural network is configured to output a tooth feature vector in response to receiving a three-dimensional digital tooth model of the one or more three-dimensional digital tooth models, and wherein the tooth model generation neural network is also configured to form a collective feature vector by connecting the tooth feature vectors of at least some of the tooth-specific context encoder neural networks.

[0180] Item 4: A dental method according to Item 3, wherein the context encoder neural network is any one of the following: a surface-based encoder, a diffusion network (DiffusionNet) encoder, a graph-based encoder, a graph neural network, a point-based encoder, a point network (PointNet) encoder, a voxel based encoder (voxel based encoder), a three-dimensional shape dataset (three-dimensional ShapeNet), a three-dimensional convolutional neural network, a multi-view rendering based neural network, a two-dimensional convolutional neural network, an implicit surface based encoder (an implicit surface based encoder)), an occupancy network and the encoder part of a convolutional autoencoder.

[0181] Item 5: The dental method of any of the preceding items, wherein one or more teeth of the subject include any of the following: one or more adjacent teeth of the replacement tooth, teeth on the opposite side of the same jaw of the replacement tooth, a group or partial dental arch, one or more opposing teeth of the replacement tooth, and combinations thereof.

[0182] Item 6: A dental method according to any of the preceding items, wherein one or more three-dimensional digital tooth models include tooth position coordinates describing an interior position of the subject's oral cavity, wherein one or more teeth of the subject include one or more adjacent teeth of the replacement tooth and / or wherein one or more teeth of the subject include one or more opposing antagonistic teeth of the replacement tooth, wherein the tooth generation neural network is further configured to output the position coordinates of the replacement tooth in response to receiving the one or more three-dimensional digital tooth models as input.

[0183] Clause 7: A dental method according to clause 6, wherein the tooth model generation neural network is configured to output the position coordinates of the replacement teeth.

[0184] Clause 8: The dental method of clause 7, wherein the decoder portion is configured to output the position coordinates. Clause 9: The dental method of clause 7, wherein the tooth model generation neural network further comprises a position prediction portion configured to output the position coordinates, and wherein the position prediction portion is configured to receive as input a concatenation of the tooth feature vectors of one or more adjacent teeth describing the tooth coordinates.

[0185] Clause 10: The dental method of clause 9, wherein the position prediction part comprises one or more connected layers. Clause 11: The dental method of any of the preceding clauses, wherein the decoder part is implemented by any of the following: a decoder of an autoencoder, a variational autoencoder decoder part, an autodecoder, a generative adversarial network, a normalized flow model, a diffusion model, and an autoregressive model.

[0186] Clause 12: A dental method according to any of the preceding clauses, wherein the latent space of the decoder part is regularized.

[0187] Item 13: A dental method according to Item 12, wherein the method further comprises iteratively modifying the generated three-dimensional digital model of the replacement tooth by using an optimization to modify the latent space vector to minimize a loss function, wherein the loss function enforces interproximal contact with one or more adjacent teeth of the replacement tooth.

[0188] Clause 14: The dental method of clause 13, wherein the loss function further enforces any of the following:

[0189] - Avoid bite contacts greater than the preset maximum depth;

[0190] - Forced contact with the opposing teeth of the replacement tooth at the preset landmarks;

[0191] - and combinations thereof.

[0192] Clause 15: A dental method according to any one of the preceding clauses, wherein the method further comprises:

[0193] - receiving surface modifications; and

[0194] - applying said surface modification to said generated three-dimensional digital model of the replacement tooth.

[0195] Clause 16: The dental method of clause 15, wherein the surface modification is received from a surface modification database. Clause 17: The dental method of clause 15, wherein the method further comprises:

[0196] - receiving an autoencoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of selected teeth of the subject into a tooth-replicating autoencoder neural network, wherein the tooth-replicating autoencoder neural network comprises an autoencoder encoding portion and an autoencoder decoding portion, wherein the autoencoder decoding portion is identical to the decoder portion of the tooth-model generating neural network; and

[0197] - determining said surface modification to deform the automatically encoded three-dimensional digital teeth model into said three-dimensional digital model of selected teeth of a subject.

[0198] Item 18: A dental method according to any one of items 15, 16 or 17, wherein the selected tooth is any one of: a tooth on the same jaw as the replacement tooth, an adjacent tooth to the replacement tooth and a tooth on the opposite jaw to the replacement tooth; and wherein if the selected tooth is a tooth on the same jaw as the replacement tooth or a tooth on the opposite jaw, the method further comprises applying a mirror transformation to the surface modification.

[0199] Clause 19: A dental method according to any one of the preceding clauses, wherein the decoder portion is configured to output the generated three-dimensional digital model of the replacement tooth by deforming a mesh of a standard shape into the generated three-dimensional digital model.

[0200] Clause 20: A dental method according to any of the preceding clauses, wherein the method further comprises: acquiring a three-dimensional digital dental structure of the subject using a three-dimensional imaging system.

[0201] Clause 21: The dental method according to clause 20, wherein the one or more three-dimensional digital tooth models are selected from the segmentation of the three-dimensional digital tooth structure using preset criteria.

[0202] Clause 22: A dental method according to any of the preceding clauses, wherein the method further comprises rendering a display of a set of one or more three-dimensional digital tooth models of the subject, including the generated three-dimensional digital model of the replacement teeth. Clause 23: A dental method according to any of the preceding clauses, wherein the method further comprises processing a dental restoration model using the generated three-dimensional digital model of the replacement teeth.

[0203] Clause 24. A dental system, wherein the dental system comprises:

[0204] - a memory storing machine executable instructions and a tooth model generating neural network, wherein the tooth model generating neural network comprises an encoder portion and a decoder portion, wherein the encoder portion is configured to output a collective feature vector describing one or more digital tooth models, wherein the tooth model generating neural network further comprises at least one fully connected layer, the fully connected layer is configured to output a latent space vector in response to receiving the collective feature vector, wherein the decoder portion is configured to output a generated three-dimensional digital model of the replacement tooth in response to receiving the latent space vector; and

[0205] - a computing system, wherein execution of the machine-executable instructions causes the computing system to:

[0206] - receiving a selection of replacement teeth for the subject;

[0207] - receiving one or more three-dimensional digital tooth models describing one or more teeth of the subject; and

[0208] - Receiving the generated three-dimensional digital model of replacement teeth in response to inputting the one or more three-dimensional digital tooth models describing one or more teeth of the subject into the tooth model generation neural network. Clause 25: A dental system according to clause 24, wherein the dental system further comprises a machining system configured to use the generated three-dimensional digital model of replacement teeth to machine a dental restoration model.

[0209] Clause 26: A computer program comprising machine executable instructions and a tooth model generating neural network, wherein the tooth model generating neural network comprises an encoder portion and a decoder portion, wherein the encoder portion is configured to output a collective feature vector describing one or more digital tooth models, wherein the tooth model generating neural network further comprises at least one fully connected layer, wherein the fully connected layer is configured to output a latent space vector in response to receiving the collective feature vector, wherein the decoder portion is configured to output a generated three-dimensional digital model of a replacement tooth in response to receiving the latent space vector, wherein execution of the machine executable instructions causes a computing system to:

[0210] - receiving a selection of replacement teeth for the subject;

[0211] - receiving one or more three-dimensional digital tooth models describing one or more teeth of the subject; and

[0212] - receiving the generated three-dimensional digital model of replacement teeth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of the subject into the tooth model generating neural network. Clause 27: A dental method, wherein the dental method comprises:

[0213] - receiving a generated three-dimensional digital model of the subject's replacement teeth;

[0214] - receiving a three-dimensional digital model of the subject's selected teeth;

[0215] - receiving an autoencoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the selected tooth of the subject into an autoencoder neural network, wherein the autoencoder neural network includes an autoencoder encoding portion and an autoencoder decoding portion;

[0216] - determining surface modifications to deform the automatically encoded three-dimensional digital tooth into a selected tooth of a subject; and

[0217] - applying said surface modification to said generated three-dimensional digital model of the replacement tooth.

[0218] Item 28: A dental method according to Item 27, wherein the selected tooth is any one of: a tooth on the same jaw as the replacement tooth, an adjacent tooth to the replacement tooth, and an opposing tooth to the replacement tooth; and wherein, if the selected tooth is a tooth on the same jaw as the replacement tooth or an opposing tooth to the replacement tooth, the method further includes applying a mirror transformation to the surface modification.

[0219] Clause 29: A dental method according to clause 27 or 28, wherein the method comprises:

[0220] - receiving a selection of replacement teeth for the subject;

[0221] - receiving one or more three-dimensional digital tooth models depicting one or more teeth of a subject;

[0222] - receiving a generated three-dimensional digital model of the replacement teeth of the subject in response to inputting one or more three-dimensional digital tooth models into a tooth model generation neural network, wherein the tooth model generation neural network includes an encoder portion and a decoder portion, wherein the decoder portion is the same as the autoencoder decoding portion, wherein the encoder portion is configured to output a collective feature vector describing one or more teeth of the subject, wherein the generation neural network also includes at least one fully connected layer, the fully connected layer is configured to output a latent space vector in response to receiving the collective feature vector, and wherein the decoder portion is configured to output the three-dimensional digital model in response to receiving the latent space vector.

[0223] Clause 30: The dental method according to clause 27 or 28, wherein the generated three-dimensional digital model of the replacement tooth is received from a database of generated three-dimensional digital tooth models.

[0224] Clause 31: A dental method according to any one of clauses 27 to 30, wherein the decoder portion is configured to generate a generated three-dimensional digital model of the replacement tooth by deforming a standard shape mesh.

[0225] Item 32. The dental method of Item 31, wherein the standard shape mesh comprises vertices, wherein the surface modification is determined by calculating a transformation of the vertices of the three-dimensional digital model of the selected tooth along normals until it intersects with the automatically encoded three-dimensional digital tooth model.

[0226] Clause 33: A dental method according to any one of clauses 27 to 31, wherein the surface modification is determined as a displacement mapping between the three-dimensional digital model of the selected tooth and the automatically encoded three-dimensional digital tooth model. Clause 34: A dental system, wherein the dental system comprises:

[0227] - a memory storing machine executable instructions; and

[0228] - a computing system, wherein execution of the machine-executable instructions causes the computing system to:

[0229] - receiving a generated three-dimensional digital model of the subject's replacement teeth;

[0230] - receiving a three-dimensional digital model of the subject's selected teeth;

[0231] - receiving an autoencoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the selected tooth of the subject into an autoencoder neural network, wherein the autoencoder neural network comprises an autoencoder encoding portion and an autoencoder decoding portion;

[0232] - determining surface modifications to deform the automatically encoded three-dimensional digital tooth into a selected tooth of the subject; and

[0233] - applying surface modifications to said generated three-dimensional digital model of the replacement tooth.

[0234] Clause 35: A dental system according to clause 34, wherein the dental system further comprises a processing system configured to process a dental restoration model using the three-dimensional digital model of the replacement tooth generated after the surface modification.

[0235] Clause 36: A computer program comprising machine-executable instructions, wherein execution of the machine-executable instructions causes a computing system to:

[0236] - receiving a generated three-dimensional digital model of the subject's replacement teeth;

[0237] - receiving a three-dimensional digital model of the subject's selected teeth;

[0238] - receiving an autoencoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the selected tooth of the subject into an autoencoder neural network, wherein the autoencoder neural network includes an autoencoder encoding portion and an autoencoder decoding portion;

[0239] - determining surface modifications to deform the automatically encoded three-dimensional digital tooth into a selected tooth of a subject; and

[0240] - applying said surface modification to said generated three-dimensional digital model of the replacement tooth.

[0241] Clause 37: A dental method, wherein the method comprises:

[0242] - receiving a generated three-dimensional digital model of a replacement tooth of a subject, wherein the generated three-dimensional digital model is formed of a mesh having a mesh topology, the mesh topology comprising vertices;

[0243] - receiving a surface modification to be transferred to a mesh forming the generated three-dimensional digital model, the surface modification being defined for the same mesh topology as the mesh forming the generated three-dimensional digital model and comprising positional adjustments to vertices of the mesh topology; and

[0244] - applying said surface modification to a mesh forming a generated three-dimensional digital model of the replacement tooth.

[0245] Clause 38: The dental method of clause 37, wherein the surface change is received from a surface modification database. Clause 39: The dental method of clause 37, wherein the method further comprises:

[0246] - receiving an autoencoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of selected teeth of the subject into a tooth-replicating autoencoder neural network, wherein the tooth-replicating autoencoder neural network comprises an autoencoder encoding portion and an autoencoder decoding portion, wherein the autoencoder decoding portion is identical to the decoder portion of the tooth-model generating neural network; and

[0247] - determining surface modifications to deform said automatically encoded three-dimensional digital teeth model into said three-dimensional digital model of selected teeth of a subject.

[0248] Item 40: A dental method according to any one of items 37, 38 or 39, wherein the selected tooth is any one of: the contralateral tooth of the replacement tooth, the adjacent tooth of the replacement tooth and the opposing tooth of the replacement tooth; and wherein the dental method further comprises, if the selected tooth is the contralateral tooth of the replacement tooth or the opposing tooth, applying a mirror transformation for surface modification.

[0249] Clause 41: A dental system, wherein the dental system comprises:

[0250] - memory storing machine-executable instructions; and

[0251] - a computing system, wherein execution of the machine-executable instructions causes the computing system to:

[0252] - receiving a generated three-dimensional digital model of a replacement tooth of a subject, wherein the generated three-dimensional digital model is formed of a mesh having a mesh topology including vertices;

[0253] - receiving a surface modification to be transferred to a mesh forming a generated three-dimensional digital model, the surface modification being directed to the same mesh topology as the mesh forming the generated three-dimensional digital model and comprising positional adjustments to mesh topology vertices; and,

[0254] - applying said surface modification to a mesh forming said generated three-dimensional digital model of a replacement tooth.

[0255] Clause 42: The dental system of clause 41, wherein the dental system further comprises a processing system configured to process a dental restoration model using the generated three-dimensional digital model of the replacement tooth.

[0256] Clause 43: A computer program comprising machine executable instructions, wherein execution of the machine executable instructions causes the computing system to: receive a generated three-dimensional digital model of a replacement tooth of a subject, wherein the generated three-dimensional digital model is formed of a mesh having a mesh topology comprising vertices;

[0257] - receiving a surface modification to be transferred to a mesh forming a generated three-dimensional digital model, said surface modification being directed to the same mesh topology as the mesh forming the generated three-dimensional digital model and comprising positional adjustments to vertices of the mesh topology; and

[0258] - applying said surface modification to a mesh forming said generated three-dimensional digital model of a replacement tooth.

[0259] Reference Number List

[0260] 50 User Interface

[0261] 52 Control elements

[0262] 54 Hardware Components

[0263] 56 Keyboard

[0264] 58 Mouse

[0265] 59 Scanner

[0266] 60 Three-dimensional printing equipment (three-dimensional printing equipment)

[0267] 62 Printing elements

[0268] 70 Manufacturing Equipment

[0269] 72 Manufacturing Tools

[0270] 74 Keep Equipment

[0271] 76 Rough

[0272] 78 raw materials

[0273] 80 Solid tooth restoration models

[0274] 100 Dental System

[0275] 102 Computer

[0276] 104 Digital Model Source

[0277] 106 Processing system

[0278] 108 Display

[0279] 110 Computing Systems

[0280] 112 Hardware Interface

[0281] 114 User Interface

[0282] 116 Memory

[0283] 120 machine executable instructions

[0284] 122 Tooth Model Generation Neural Network

[0285] 124 Choice of replacement teeth

[0286] 126 One or more 3D digital tooth models

[0287] 128 Generated 3D digital model of replacement teeth 128' The generated three-dimensional digital model with position coordinates of the replacement tooth

[0288] 130 tooth model generation neural network collection

[0289] 132 Position coordinates of the generated three-dimensional digital model of the replacement tooth

[0290] 134 Optimization Algorithm

[0291] 136 Surface modification

[0292] 138 3D digital model of replacement tooth with optional surface modification

[0293] 140 Computer-Aided Design Module

[0294] 142 Control Commands

[0295] 150 A set of one or more 3D digital tooth models

[0296] 300 Dental Systems

[0297] 302 3D digital model of selected teeth

[0298] 304 Autoencoder Neural Network

[0299] 306 Automatically encoded 3D digital tooth model

[0300] 136 Surface modification

[0301] 138 Surface-modified 3D digital model of replacement tooth

[0302] 200 Receive the subject's choice of replacement teeth 202 Receiving one or more three-dimensional digital tooth models describing one or more teeth of a subject

[0303] 204 Using the selection of replacement teeth, a tooth model generation neural network for generating replacement teeth is selected from a set of tooth model generation neural networks trained for specific teeth or specific tooth types.

[0304] 206 Receiving the generated three-dimensional digital model of replacement teeth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of the subject into the tooth model generation neural network

[0305] 208 Iteratively modify the generated three-dimensional digital model of the replacement tooth by modifying the potential space vector by optimizing and minimizing the loss function

[0306] 210 Output the position coordinates of the replacement tooth

[0307] 212 Receive surface modification

[0308] 214 Apply the surface modification

[0309] 216 Rendering and displaying a set of one or more three-dimensional digital tooth models of the subject including the generated three-dimensional digital model of the replacement teeth 218 Using the generated three-dimensional digital model of the replacement tooth to manufacture a dental restoration 308 3D digital tooth model library

[0310] 310 Surface modification database

[0311] 400 Receive the generated 3D digital model of the replacement tooth

[0312] 402 Receive the three-dimensional digital model of the selected tooth 404 Receiving an auto-encoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of the selected tooth into the auto-encoder neural network 406 Determine the surface modification that can deform the automatically encoded three-dimensional digital tooth model into the three-dimensional digital model of the selected tooth

[0313] 600 3D digital tooth structure

[0314] 602 Segment

[0315] 604 Three-dimensional digital tooth model

[0316] 606 Gums

[0317] 700 Missing Teeth

[0318] 800 Input Grid

[0319] 802 Output Grid

[0320] 804 Encoder Section

[0321] 806 decoder part

[0322] 808 Middle adjacent teeth

[0323] 810 Distal Neighboring Teeth

[0324] 812 Opposite teeth in the same jaw

[0325] 814 Tooth-specific context encoder

[0326] 816 tooth feature vectors

[0327] 818 Collective eigenvector

[0328] 820 fully connected layers

[0329] 822 Latent space vector

[0330] 824 Standard shapes

[0331] 826 Position prediction part

[0332] 827 Additional eigenvectors

[0333] 828 fully connected layers

[0334] 830 Location coordinates

[0335] 840 Autoencoder Neural Network

[0336] 842 Autoencoder encoding part

[0337] 844 Autoencoder decoding part

[0338] 846 3D digital model of selected teeth

[0339] 848 Automatically encoded 3D digital tooth model

Claims

1. A dental method, wherein the method comprises: - Receive the subject's selection of replacement teeth; - receiving one or more three-dimensional digital tooth models depicting one or more teeth of a subject; and - receiving a generated three-dimensional digital model of replacement teeth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of a subject into a tooth model generating neural network, wherein the tooth model generating neural network includes an encoder portion and a decoder portion, wherein the encoder portion is configured to output a collective feature vector describing the one or more digital tooth models, wherein the tooth model generating neural network also includes at least one fully connected layer, which is configured to output a latent space vector in response to receiving the collective feature vector, and wherein the decoder portion is configured to output the generated three-dimensional digital model of replacement teeth in response to receiving the latent space vector.

2. The dental method of claim 1, wherein the method further comprises using the selection of replacement teeth to select a tooth model generation neural network for generating the replacement teeth from a set of tooth model generation neural networks trained for specific teeth or specific tooth types.

3. The dental method according to claim 1, wherein the encoder portion comprises a tooth-specific context encoder neural network for each of the one or more three-dimensional digital tooth models, wherein the tooth-specific context encoder neural network is configured to output a tooth feature vector in response to receiving a three-dimensional digital tooth model of the one or more three-dimensional digital tooth models, and wherein the tooth model generating neural network is further configured to form a collective feature vector by connecting the tooth feature vectors of at least some of the tooth-specific context encoder neural networks.

4. The dental method according to claim 3, wherein the context encoder neural network is any one of the following: a surface based encoder, a diffusion network (DiffusionNet) encoder, a graph based encoder, a graph neural network, a point based encoder, a point network (PointNet) encoder, a voxel based encoder (voxel based encoder), a three-dimensional shape dataset (three-dimensional ShapeNet), a three-dimensional convolutional neural network, a multi-view rendering based neural network, a two-dimensional convolutional neural network, an implicit surface based encoder (implicit surface based encoder)), an occupancy network and the encoder part of a convolutional autoencoder.

5. The dental method according to claim 1, wherein the one or more teeth of the subject include any of the following: one or more adjacent teeth of the replacement tooth, teeth on the opposite side of the same jaw of the replacement tooth, a group or partial dental arch, one or more opposing teeth of the replacement tooth, and combinations thereof.

6. The dental method according to claim 1, wherein the one or more three-dimensional digital tooth models include tooth position coordinates describing the internal position of the subject's oral cavity, wherein the one or more teeth of the subject include one or more adjacent teeth of the replacement tooth and / or wherein the one or more teeth of the subject include one or more opposing teeth of the replacement tooth, wherein the tooth generation neural network is further configured to output the position coordinates of the replacement tooth in response to receiving the one or more three-dimensional digital tooth models as input.

7. The dental method according to claim 6, wherein the tooth model generation neural network is configured to output the position coordinates of the replacement tooth.

8. The dental method of claim 7, wherein the decoder portion is configured to output the position coordinates.

9. The dental method according to claim 7, wherein the tooth model generation neural network further comprises a position prediction part configured to output the position coordinates, and wherein the position prediction part is configured to receive as input a connection of the tooth feature vectors of one or more adjacent teeth describing the tooth coordinates.

10. The dental method of claim 9, wherein the position prediction portion comprises one or more connected layers.

11. The dental method according to claim 1, wherein the decoder part is implemented by any one of the following: a decoder of an autoencoder, a variational autoencoder decoder part, an autodecoder, a generative adversarial network, a normalized flow model, a diffusion model, and an autoregressive model.

12. The dental method according to claim 1, wherein the latent space of the decoder part is regularized.

13. The dental method of claim 12, wherein the method further comprises iteratively modifying the generated three-dimensional digital model of the replacement tooth by modifying the latent space vector using optimization to minimize a loss function, wherein: The loss function enforces interproximal contact with one or more adjacent teeth of the replacement tooth.

14. The dental method according to claim 13, wherein: The loss function further enforces either: - Avoid bite contacts greater than the preset maximum depth; - Forced contact with the opposing teeth of the replacement tooth at the preset landmarks; - and combinations thereof.

15. The dental method of claim 1, wherein the method further comprises: -Receive surface modifications; and - applying said surface modification to said generated three-dimensional digital model of the replacement tooth.

16. The dental method according to claim 15, wherein: The surface modifications are received from a surface modification database.

17. The dental method of claim 15, wherein the method further comprises: - receiving an autoencoded three-dimensional digital tooth model in response to inputting the three-dimensional digital model of selected teeth of the subject into a tooth-replicating autoencoder neural network, wherein the tooth-replicating autoencoder neural network comprises an autoencoder encoding portion and an autoencoder decoding portion, wherein the autoencoder decoding portion is identical to the decoder portion of the tooth-model generating neural network; and - determining said surface modification to deform the automatically encoded three-dimensional digital teeth model into said three-dimensional digital model of selected teeth of a subject.

18. A dental method according to claim 15, wherein the selected tooth is any one of the following: the contralateral tooth of the replacement tooth, the adjacent tooth of the replacement tooth and the opposing tooth of the replacement tooth; and wherein if the selected tooth is the contralateral tooth of the replacement tooth or the opposing tooth, the method also includes applying a mirror transformation to the surface modification.

19. The dental method according to claim 1, wherein the decoder portion is configured to output the generated three-dimensional digital model by deforming a mesh of a standard shape into the generated three-dimensional digital model of the replacement tooth.

20. The dental method of claim 1, wherein the method further comprises: A three-dimensional imaging system was used to obtain the three-dimensional digital dental structure of the subject.

21. The dental method according to claim 20, wherein: The one or more three-dimensional digital tooth models are selected from the segmentation of the three-dimensional digital tooth structure using preset criteria.

22. The dental method of claim 1, wherein the method further comprises rendering a display of a set of one or more three-dimensional digital tooth models of the subject, including the generated three-dimensional digital model of the replacement tooth.

23. The dental method of claim 1, further comprising machining a dental restoration model using the generated three-dimensional digital model of the replacement tooth.

24. A dental system, wherein: The dental system comprises: - a memory storing machine executable instructions and a tooth model generating neural network, wherein the tooth model generating neural network comprises an encoder portion and a decoder portion, wherein the encoder portion is configured to output a collective feature vector describing one or more digital tooth models, wherein the tooth model generating neural network further comprises at least one fully connected layer, the fully connected layer is configured to output a latent space vector in response to receiving the collective feature vector, wherein the decoder portion is configured to output a generated three-dimensional digital model of the replacement tooth in response to receiving the latent space vector; and - a computing system, wherein execution of the machine-executable instructions causes the computing system to: - receiving a selection of replacement teeth for the subject; - receiving one or more three-dimensional digital tooth models describing one or more teeth of a subject; and - receiving a generated three-dimensional digital model of a replacement tooth in response to inputting the one or more three-dimensional digital tooth models describing one or more teeth of the subject into the tooth model generating neural network.

25. The dental system of claim 24, wherein the dental system further comprises a machining system configured to machine a dental restoration model using the generated three-dimensional digital model of the replacement tooth.

26. A computer program comprising machine executable instructions and a tooth model generating neural network, wherein the tooth model generating neural network comprises an encoder portion and a decoder portion, wherein the encoder portion is configured to output a collective feature vector describing one or more digital tooth models, wherein the tooth model generating neural network further comprises at least one fully connected layer, wherein the fully connected layer is configured to output a latent space vector in response to receiving the collective feature vector, wherein the decoder portion is configured to output a generated three-dimensional digital model of a replacement tooth in response to receiving the latent space vector, wherein execution of the machine executable instructions causes a computing system to: - receiving a selection of replacement teeth for the subject; - receiving one or more three-dimensional digital tooth models describing one or more teeth of the subject; and - receiving said generated three-dimensional digital model of replacement teeth in response to inputting one or more three-dimensional digital tooth models describing one or more teeth of a subject into said tooth model generating neural network.

Images